SOCIOPHONETICS OF THE LE HAVRE ACCENT

Damien Hall

Newcastle University (UK) [email protected]

ABSTRACT 24, 28, 33, 35, 36, 43, 47], the dominant perception is that variation in French across Northern is We report on a pilot analysis of two speakers—M, 33, mostly lexical [4]. and F, 24, both middle-class—from Le Havre, The French spoken in Le Havre is of interest France, part of the larger Towards A New Linguistic because it is subject to two possibly contradictory Atlas of France project.1 The aim is to isolate features influences. On the one hand, France has been to investigate in greater detail in the full analysis. characterised as a ‘hypercephalic’ country [2], where Two vowel changes are analysed: the merger or the capital has nationwide influence out of proportion separation of /a/ (as in patte /pat/ ‘paw’) and /ɑ/ (pâtes to the distance between it and the country’s other /pɑt/ ‘pasta’), and the fronting of /ɔ/. Most areas of large urban centres. Le Havre is relatively close to France merge /a/ and /ɑ/ to /a/ [18, 24, 46], but some (215km/134mi), and the two are connected by Normandy speakers separate them [21], as does the road, rail and river, so we would expect Paris to be a Norman, the oral vowel system of significant influence on it anyway, but the influence which is very close to that of its sister language is still magnified by the organisation of the country. French [30]. Both speakers analysed here have On the other hand, the region of Normandy has a significant word-list differences between /a/ and /ɑ/. well-known autochthonous variety, Norman, a sister The female speaker also has fronted /ɔ/, a well-known variety to French [30]. Norman was once spoken over feature of modern informal French [3], but one which a large area from Dieppe in the East to the Channel has not been found for Norman. The emerging picture Islands in the West, but it now has no monolingual is of an urban accent which combines regional speakers, and relatively few bilingual ones [30]. features and more widespread urban ones, even There are some adult learners of Norman in both among middle-class speakers. mainland France and (especially) the Channel Islands [29]. Despite this apparent moribundity, this paper Keywords: sociophonetics, phonetics, phonology, finds a possible influence from Norman in the French French, Norman of Le Havre. We analyse word-list productions of two pairs of 1. INTRODUCTION vowels, the /a ~ ɑ/ pair and the /ɔ ~ œ/ pair. /a/ and /ɑ/ are phonemes of classical French, and also of modern This paper presents initial results from the ongoing , though much of the French of Towards A New Linguistic Atlas of France project France (at least from middle-class speakers) now (TANLAF; [22]). TANLAF aims to investigate merges them to /a/ [11, 21, 45]. Despite this, a possible differences between the distinction which is at least allophonic, if not spoken in the major towns and cities of Northern phonemic, has been found in the Regional ; we present here results from Le Havre, one of Normandy in sites other than Le Havre [21], and the the largest cities (population approx. 300,000) in the distinction in Norman is also at least allophonic [30]. largely rural region of Normandy. The relationship between /ɔ/ and /œ/ is of a The dominant wisdom in the field is that there are different kind: fronting of /ɔ/ towards [œ] has been not many distinct regional varieties of French within noted since at least the late 1950s [38]. It still seems the North of France (the langue d’oïl area, where the to be productive in much urban French of the North autochthonous varieties are closely related to of France [1], and there is evidence it is now ) [2, 11]. Native speakers themselves spreading to rural areas too [38]. The study in [9] is also find it difficult to tell apart the varieties of French divided into Northern French and Southern French spoken by natives of different cities in the North of samples, and each sample includes both urban and France [7, 8], though the area around the far Northern rural speakers, indicating that they did not expect to city of Lille may be an exception [41, 42]. Overall, find much difference between urban and rural despite recent work showing large numbers of treatments of /ɔ/ fronting. Our analysis therefore speakers throughout France pronouncing certain aimed to see whether it was also present in Le Havre. words in non-normative ways [4], and work showing phonological variation and change in Paris [12, 13,

2645

2. METHOD midpoints, using a modified version of [34]. Reliability was ensured because recommended use of We analyse two of the TANLAF Le Havre EasyAlign requires checking phone boundaries; this informants: a 24-year-old woman (F24) and a 33- also gave the opportunity to check signal quality and year-old man (M33). Following [32], final sample the reliability of Praat’s formant detection. size per city will vary with the size of the city; as Le Raw formant frequencies in Hz were Lobanov- Havre has between 200,000 and 1,000,000 normalised through phonR [39]. The potential inhabitants, its final sample will be two men and two presence of distinctions between all relevant vowel women. Within each city, the sample is stratified only pairs for each speaker was then tested by t-test on the by biological sex. Informants must have spent the normalised F1 (height) and F2 (anteriority) values. majority of their life in the urban area of the city concerned, especially since the age of 4, so that most 3. RESULTS of their peers are from there [31]. All informants are aged 18-33, middle-class, with at most one university Raw Hz formant frequencies, normalised values, and degree, and none are educators. These criteria are in full vowel-plots for F24 and M33 are provided at order to eliminate possible confounding motivations [23]. for variation. In this way, any variation found The partial vowel-plots in Figures 1 and 2 show between cities is more likely to be regional variation, the relationships between /a/ and /ɑ/, and /ɔ/, /œ/ and and not caused by social class differences, age /ə/, for F24 and M33. The figures compare the mean differences or stereotypes which may dictate the kind values of the speakers’ vowels with a reference set of of French that an educated person ‘should’ speak. vowels, mostly from [15], but including /ə/ from [49], Recordings are made in relaxed surroundings as [15] does not include it. (these informants were recorded in their homes). The speakers analysed here were recorded on a Marantz Figure 1: Low and mid vowels of speaker F24 PMD671 solid-state recorder, recording direct to compared with female reference vowels [15]. .wav (sampling rate of 22.05kHz / 16-bit). As storage of large files is now much easier than it was, many studies now record at double this rate, but 22.05kHz is also still used in studies of vowels [e.g. 5, 44], as it provides reliable measurements for frequencies below 11.025kHz [16], which is still far above the frequencies which have been found to be sociolinguistically relevant for vowels, up to about 4kHz [25]. The microphone was a collar-mounted Audio-Technica PRO70 cardioid condenser lavalier. Speakers recorded an interview with the researcher, a reading passage and a word-list. Word- list data is analysed in this pilot study, on the basis that any regional difference appearing there will certainly appear in less formal connected speech. The word-list was presented in PowerPoint, one word per Figure 2: Low and mid vowels of speaker M33 slide, so as not to show speakers when the end of the compared with male reference vowels [15]. list was approaching. Analysis was done in Praat [6]. Table 1 shows the tokens analysed here.

Table 1: Tokens analysed in this paper

/a/ /ɑ/ /ɔ/ /œ/ /ə/ F24 87 32 62 18 20 M33 120 26 70 25 144

The discrepancy between speaker totals is explained by the removal of outliers and badly- recorded tokens. Recordings were segmented using EasyAlign [20]. Formant values were measured at vowel

2646

We use a relatively old reference set because more Table 3: Differences between /ɔ/ and /œ/ modern reference sets ([19, 49]) do not include /ɑ/: most speakers of French in France, particularly young Height Anteriority df and middle-aged ones, do not now make an audible t p t p difference between /a/ and /ɑ/ [11, 17]. F24 79 3.345 <0.01 9.999 <0.001 M33 94 4.227 <0.001 11.705 <0.001 3.1. /a/ and /ɑ/ Table 4: Differences between /ɔ/ and /ə/ Tokens were coded as /ɑ/ if their phonological Height Anteriority environment was listed in [14] as being likely to df produce /ɑ/ and not /a/. A list of phonological t p t p environments for was used in this F24 81 1.908 >0.05 2.369 <0.05 study of the French of France because Quebec French M33 213 6.087 <0.001 8.573 <0.001 still reliably distinguishes /ɑ/ from /a/ in the way that the French of France once did, though the distinction /ɔ/ and /œ/ should of course be significantly is now rare in the French of France. Token numbers different for any speaker, but the test was carried out for /ɑ/ are much lower than numbers for /a/ simply as a sense-check, and gave the expected result (Table because of the low frequency of /ɑ/ relative to /a/ in 3). Of more interest are the differences between /ɔ/ French of any variety (for speakers with a distinction) and /ə/, since fronted /ɔ/ would be closer to /ə/ than to or of the relevant phonological environments (for /œ/. Therefore, in straightforward /ɔ/-fronting, there speakers without one). might not be any significant difference between a Table 2 shows the results of t-tests on the speaker’s /ɔ/ and /ə/ at all. However, if there were any differences between /a/ and /ɑ/ for our speakers. For significant difference, one might expect it to be in both F24 and M33, /a/ is significantly higher than /ɑ/. height rather than in anteriority, since normative /ə/ is In addition, /a/ is significantly fronter than /ɑ/ for higher than normative /ɔ/ [17]. The full version of M33, though this is not true for F24. F24’s vowel chart [23] shows P. Durand (1985)’s reference /ə/ much higher than reference /ɔ/, and also Table 2: Differences between /a/ and /ɑ/ much higher than F24’s /ə/ and /ɔ/. In fact, for F24, the differences are the other way Height Anteriority around from what we might expect given the norm. df t p t p /ɔ/ is significantly different from /ə/ in anteriority, but F24 118 2.107 <0.05 0.249 >0.05 not in height. The significant anteriority difference M33 145 3.591 <0.01 2.960 <0.01 looks on the face of it as if we should reject the hypothesis that F24 is fronting /ɔ/. In terms of 3.2. /ɔ/, /œ/ and /ə/ absolute (Euclidean) distance, though, F24’s /ɔ/ is slightly closer to her own /ə/ than it is to reference /ɔ/ The relationships tested among these three vowels (Table 5). Therefore, taking into account that the were between /ɔ/ and /œ/ and between /ɔ/ and /ə/. In phonetic effect of /ɔ/-fronting is to move /ɔ/ closer to the Standard French vowel-space [17], /ɔ/ is a mid- /ə/, we can still conclude that F24 is marginally low rounded back vowel and /œ/ a mid-low rounded participating in the /ɔ/-fronting change in progress. front vowel, both at about the same height, with /ə/ a For M33, /ɔ/ and /ə/ are highly significantly mid-central vowel a little higher than these. When /ɔ/ separated in both height and anteriority: he does not is fronted, it typically does not front further than a seem to be participating in /ɔ/-fronting, at least in central position, so that it ends closer to /ə/ than to formal word-list style. /œ/. Nevertheless, the prototypical symbol associated with fronted /ɔ/ is <œ> [38, 1, 40], because this is how Table 5: Euclidean distances for vowels French often spells both /ə/ and /œ/. Tables 3 and 4 involved in /ɔ/-fronting show the results of t-tests on these differences. speaker /ɔ/- speaker /ɔ/- speaker /ə/ reference /ɔ/ F24 0.459 0.521 M33 0.775 0.508

2647

4. DISCUSSION and /ɑ/ has also been observed in cities, particularly the suburbs of Paris [28], the urban area near Lille The sociophonetic variation investigated in this paper [27] and Le Havre itself [26]. /ɔ/-fronting is more covers two different types of change in French: unambiguously an incoming feature, and it is not 1) a phonological change, between a system with surprising that our female speaker should exhibit it two low unrounded vowels, /a/ and /ɑ/, as in while our male speaker does not: it is axiomatic in more conservative varieties, and a system with sociolinguistics that young female speakers often lead one, /a/, as in less conservative varieties in the adoption of new linguistic changes [31]. 2) a phonetic change, the ongoing sound-change Thus, our young female speaker can be said to of /ɔ/-fronting, which implicates the exhibit at least two elements of non-standard, urban relationship between a speaker’s /ɔ/ and a French, while our young male speaker can be said not speaker’s /ə/ (and reference /ɔ/), but does not to exhibit any, even though their treatment of one of implicate their number of vowel phonemes. the variables examined here is the same. This apparent paradox hinges on the status of an /a ɑ/ 4.1. /a/ and /ɑ/ separation as both a feature of conservative Standard French and a feature of (Northern) urban non- In conservative French [50], we expect /a/ higher and standard French. It does not seem necessary to resolve fronter than /ɑ/. We find this configuration for M33. the paradox: one linguistic feature can perfectly well F24, on the other hand, has /a/ significantly higher but be part of more than one system. Confirmation of not significantly fronter than /ɑ/. It is interesting that other non-standard characteristics of a Le Havre the conservative difference reflected in descriptions accent will await further investigation of the data like [50] is not reflected in datasets like [15]: in [15], collected, and future studies of the city using a bigger in fact, /ɑ/ is slightly higher and slightly fronter than sample size. We can now say, though, that there may /a/, the opposite of most descriptions which have both be more urban variation in France than we have vowels (though not all descriptions have both). thought (cf [2] and studies summarised there). We can Significance cannot be tested for the vowels in [15], also challenge the characterisation in [26] of the Le as only one male and one female value are given. Havre accent as a ‘linguistic myth’. Perceptually, we should note that, even in Normandy speakers who have /a/ and /ɑ/ significantly 5. REFERENCES different, the difference is hard to hear, at least consciously [21]. This raises the questions of how [1] Armstrong, N., Low, J. 2008. C’est encœur plus jeuli, relevant Normandy’s statistically significant le Mareuc: some evidence for the spread of /ɔ/- difference is to perception and to acquisition. These fronting in French. Trans. Phil. Soc. 106(3), 432-455. [2] Armstrong, N., Pooley, T. 2010. Social and linguistic questions are sociolinguistic, not phonetic, but they change in . New York / could be tested by future perceptual experiments. Basingstoke: Palgrave Macmillan. [3] Armstrong, N., Pooley, T. 2013. Levelling, resistance 4.2. /ɔ/, /œ/ and /ə/ and divergence in the pronunciation of English and French. Language Sciences 39, 141-150. /ɔ/-fronting is described in the literature as a trait of [4] Avanzi, M. 2017. Atlas du français de nos régions. Northern urban French [1, 26, 38, 40]. F24 is Malakoff: Armand Colin. participating in this ongoing change, while M33 is https://francaisdenosregions.com/ not; in fact, M33’s /ɔ/ is slightly backed and raised [5] Blackwood Ximenes, A., Shaw, J.A., Carignan, C. compared to reference /ɔ/ (Figure 2). 2017. A comparison of acoustic and articulatory methods of analysing vowel differences across

. J. Acoust. Soc. Am. 142, 363-377. 4.3 Conclusion [6] Boersma, P., Weenink, D. 1992-2016. Praat. Version 6.0.14. http://www.praat.org In summary, both F24 and M33 have significant [7] Boughton, Z. 2005. Accent levelling and accent separations (in at least one dimension) between /a/ localisation in northern French: Comparing Nancy and /ɑ/, and F24 is participating in /ɔ/-fronting, and Rennes. J. French Lang. Stud. 15, 235-256. though M33 is not. These indications suggest that [8] Boughton, Z. 2006. When perception isn’t reality: both speakers analysed here have elements of a Le Accent identification and perceptual dialectology in Havre accent. A separation of /a/ and /ɑ/ is a trait of French. J. French Lang. Stud. 16, 277-304. Standard French in at least some descriptions [10, 15, [9] Boula de Mareüil, P., Adda-Decker, M., Woehrling, C. 2010. Antériorisation/aperture des voyelles /ɔ /~/o/ 37, 48], although many recent sources also say that en français du Nord et du Sud. 28es Journées d’Étude the distinction is now at best tenuous [18, 24, 46]. sur la Parole, Mons, 81-84. More interestingly for our purposes, use of both /a/

2648

[10] Delattre, P. 1951. Principes de phonétique française [30] Jones, M.C. 2014. Variation and Change in Mainland à l’usage des étudiants anglo-américains. and Insular Norman: a study of superstrate Middlebury, VT: Middlebury College École relationships. Leiden / Boston: Brill. Française d’Été. [31] Labov, W. 2001. Principles of Linguistic Change: [11] Detey, S., Durand, J., Laks, B., Lyche, C. 2010. Les social factors. Malden / Oxford: Blackwell. variétés du francais parlé dans l’espace francophone. [32] Labov, W., Ash, S., Boberg, C. 2006. The Atlas of Paris: Ophrys. North American English. Berlin: Mouton de Gruyter. [12] Deyhime, G. 1967a. Enquête sur la phonologie du [33] Landick, M. 1995. The mid-vowels in figures: hard français contemporain. La Linguistique 3(1), 97-108. facts. French Review 69(1), 88-103. [13] Deyhime, G. 1967b. Enquête sur la phonologie du [34] Lennes, M. 2003. français contemporain. La Linguistique 3(2), 57-84. collect_formant_data_from_files.praat. [14] Dumas, D. 1986. Le statut des deux ‘A’ en français https://github.com/lennes/spect/tree/master/scripts québécois. Revue Québécoise de ling. 15(2), 167-197. [35] Lennig, M. 1978. Acoustic Measurement of [15] Durand, P. 1985. Variabilité acoustique et invariance Linguistic Change: the modern Paris vowel system. en français. Paris: CNRS. PhD thesis, University of Pennsylvania. Distributed [16] Ellis, D.P.W. 2010. An introduction to signal as Pennsylvania Dissertation Series, no. 1. processing for speech. In: Hardcastle, W.J., Laver, J., [36] Léon, P. 1972. Étude de la pronunciation du ‘e’ Gibbon, F.E. The Handbook of Phonetic Sciences. accentué chez un groupe de jeunes Parisiens. In: Malden / Oxford / Chichester: Wiley, 757-780. Valdman, A. (ed), Papers in Linguistics and [17] Fougeron, C., Smith, C.L. 1999. French. In: Int’l Phonetics to the Memory of Pierre Delattre (Ianua Phonetic Association. Handbook of the International Linguarum 54). The Hague: Mouton. Phonetic Association. Cambridge: CUP, 78-81. [37] Martinet, A. 1945. La Prononciation du Français [18] Gadet, F. 1989. Le français ordinaire. Paris: Armand Contemporain. Paris: Droz. Colin. [38] Martinet, A. 1957. ‘C’est jeuli, le Mareuc!’ Romance [19] Gendrot, C., Adda-Decker, M. 2005. Impact of Philology 11(4), 345-355. duration on F1/F2 values of oral vowels. Proc. [39] McCloy, D. 2016. phonR. Version 1.0.7. Interspeech 2005 Lisbon, 2453-2456. http://drammock.github.io/phonR [20] Goldman, J.-P. 2012. EasyAlign: phonetic alignment [40] Mooney, D. 2015. ‘C’est jeuli, la Gasceugne!’: with Praat. l’antériorisation du phoneme /ɔ/ dans le français http://latlcui.unige.ch/phonetique/easyalign.php régional du Béarn. French Studies 70(1), 61-81. [21] Hall, D. 2008. A Sociolinguistic Study of the Regional [41] Pooley, T. 1996. Chtimi: The urban vernaculars of French of Normandy Unpublished PhD thesis, Northern France. Clevedon: Multilingual Matters. University of Pennsylvania. [42] Pooley, T. 2004. Language, and Identity in [22] Hall, D. 2013. The Linguistic Geography of the Lille. Lewiston: Edwin Mellen Press. French of Northern France. Language and Linguistics [43] Scherrer, Y., Boula de Mareüil, P., Goldman, J.-P. Compass 7/9, 477-499. 2015. Crowdsourced mapping of pronunciation [23] Hall, D. 2019. Dataset for ‘Sociophonetics of the Le variants in European French. Proc. 18th ICPhS Havre accent’. Glasgow, 1-5. https://doi.org/10.5281/zenodo.2616604 [44] Strycharczuk, P., Scobbie, J.M. 2017. Fronting of [24] Hansen, A.B., Juillard, C. 2011. Le phonologie Southern British high-back vowels in articulation and parisienne à trente ans d’intervalle—Les voyelles à acoustics. J. Acoust. Soc. Am. 142, 322-331. double timbre. J. French Lang. Stud. 21, 313-359. [45] Walker, D. 1984. The Pronunciation of Canadian [25] Harrington, J. Acoustic Phonetics. In: Hardcastle, French. Ottawa: U. Ottawa Press. W.J., Laver, J., Gibbon, F.E. The Handbook of [46] Walter, H. 1977. La phonologie du français. Paris: Phonetic Sciences. Malden / Oxford: Wiley, 81-129. PUF. [26] Hauchecorne, F., Ball, R. 1997. L’accent du Havre: [47] Walter, H. 1992. Les fluctuations mettent-elles en un exemple de mythe linguistique. Langage et société danger une opposition phonologique? La 82, 5-25. Linguistique 28(1), 59-68. [27] Hornsby, D. 2006. Redefining regional French. [48] Warnant, L. 1987. Dictionnaire de la prononciation London: Legenda. française dans sa norme actuelle. Paris: Duculot. [28] Jamin, M. 2007. The ‘Return’ of [ɑ] in Parisian [49] Woehrling, C. 2009. Accents régionaux en français: French: a case of sociolinguistic recycling? perception, analyse et modélisation à partir de Nottingham French Stud. 46(2), 23-39. grands corpus. Unpub. PhD thesis, U. Paris-Sud. [29] Jones, M.C. 2001. Norman French: a [50] Zink, G. 1986. Phonétique Historique du Français. linguistic study of an obsolescent dialect. Oxford / Paris: PUF. Boston: Blackwell.

1 The research reported here was funded by a Leverhulme willing to give their data, and also Caitlin Halfacre, for Trust Early Career Fellowship held by the author (ECF- her help with Figures 1 & 2. Remaining shortcomings are 2010-0556). I thank them, the informants for being of course my own.

2649